LeituraPesquisaRadarFramework de investimento
Entrar / Cadastrar
TudoTecnologiaProdutosInvestimentoBolsaOutros

Leitura

2026-10-22168 posts

Atualizar

UK Bank Barclays Announces 5,000 Job Cuts Citing AI Efficiency Gains

Tópico · 巴克莱银行AI裁员Material
2026-10-22 16:38 GMT+8

British bank Barclays announced on October 22, 2026, that it will cut 5,000 jobs, representing approximately 5% of its global headcount, as part of a restructuring driven by increased efficiency from artificial intelligence.

The majority of the cuts will affect operational and functional roles within the UK retail banking arm, though investment banking will also see reductions. So far, 1,000 positions have been closed, with the remaining 4,000 expected to be eliminated before the end of the year.

Chief Executive CS Venkatakrishnan stated that the company has been making "deep investments" in technology and process simplification, which are driving most of the current efficiency gains. He described AI as playing an important role and labeled it the "icing on the cake" of these technological strategies.

Chief Financial Officer Anna Cross explained that the cuts are part of a plan to save the company £1.1 billion ($1.4 billion) in 2027. Approximately 60% of the reductions relate to support functions, with Cross noting, "AI makes our people more productive."

Fontes:TechRadar

Pesquisa

DeepSeek releases open-source inference framework Deepinfer to cut self-hosting costs

Tópico · DeepSeek推理框架Material
2026-10-21 22:00 GMT+8

Chinese AI company DeepSeek released an open-source inference framework named Deepinfer on GitHub on October 21, 2026. According to a report by German tech outlet t3n, the tool aims to help enterprises that deploy and run large language models themselves to lower their computing and energy costs, making AI usage more affordable.

Deepinfer optimizes the inference process through three main technical approaches: distributing computational tasks more evenly across available processor cores; utilizing 'Speculative Decoding' to speed up response times; and better structuring model outputs before final delivery. Currently, the framework only supports DeepSeek's own V4.1-Flash model, with plans to expand to other proprietary models in the future.

It is important to note that Deepinfer is not intended for general consumers or web-based users, but rather for enterprise clients with high-performance server clusters. It requires underlying hardware equipped with Nvidia GPUs based on the SM100 architecture or Huawei's Ascend 950 NPUs. These chips are typically offered as part of complete server systems to enterprises and are expensive; for instance, a single Huawei Ascend 950 chip costs over $37,000. Consequently, this tool primarily serves large tech companies or cloud service providers with significant self-hosting capabilities.

Fontes:t3n.de

Pesquisa

German Software Giant SAP Reports Record Quarterly Revenue as AI Strategy Drives Cloud Growth

Tópico · SAP业绩Material
2026-10-22 09:01 GMT+8

SAP SE, Europe’s largest software company, reported a record-breaking third quarter on October 21, 2026. Total sales reached €10.27 billion (approximately $11.2 billion), marking the first time the company has surpassed the €10 billion threshold in a single quarter. This figure outperformed analyst expectations of €10.10 billion. Adjusted earnings per share came in at €1.89, beating the consensus estimate of €1.83.

The strong performance of its cloud business was the primary engine behind this new revenue record. SAP’s third-quarter cloud revenue increased by 25% year-over-year to €6.67 billion. A more forward-looking metric, the current cloud backlog (contracted but unrecognized revenue), surged 26% to €24.11 billion, exceeding the 24% to 25% growth range forecast by analysts. Non-IFRS operating profit also rose to €3.03 billion, up 18% from the same period a year earlier.

Chief Executive Christian Klein explicitly stated during the earnings call that “AI was the main driver behind customer’s move to the cloud during the quarter.” This assertion aligns with the company’s recently promoted “autonomous enterprise” strategy. A key component of this strategy is Joule, an AI agent tool designed to act as a conversational interface layer connecting to over 70 AI models, working alongside rather than just within existing SAP applications. Klein revealed that hundreds of customers have signed up for new subscriptions to access Joule.

Despite the impressive results, SAP management remains cautious about the future economic environment. Chief Financial Officer Dominik Asam noted that the fourth quarter will be critical in determining how the rest of the financial year lands. He warned that the company is operating amid a gloomy economic backdrop, characterized by rising oil prices and U.S. interest rates that are “starting to develop teeth.” Asam also mentioned that part of the quarter’s cloud growth was driven by customers returning from “financial distress.”

Regarding guidance, SAP maintained its full-year operating profit outlook of €11.8 billion to €12.2 billion, which was first revealed in July. For fiscal 2026, Asam expects subscription revenue to reach the upper end of the forecast range between €25.8 billion and €26.2 billion, representing growth of 23% to 25%. Taking into account the decline of its classic perpetual license business, full-year revenue should fall between €36.3 billion and €36.8 billion, up by 12% to 13%.

Fontes:SiliconANGLE

Pesquisa

Street Signals Podcast: Orix Life CIO Shintaro Yoshida on Capital Repatriation Amid Japan's Rate Normalization

Tópico · 日本资本回流Material
2026-10-22 12:55 GMT+8

State Street's Street Signals podcast released a new episode titled "Yen, Rates and Capital Flows: The View from Tokyo" on October 22, 2026. The guest is Shintaro Yoshida, Chief Investment Officer (CIO) of Orix Life Insurance. As a decision-maker at a Japanese life insurer managing one of the world's largest pools of foreign savings, his asset allocation choices significantly impact global bond markets and the yen.

According to the publisher's show notes, the episode focuses on whether the story of persistent yen weakness is starting to change amid the normalization of domestic interest rates in Japan. Specific topics include how rising Japanese Government Bond (JGB) yields are reshaping the domestic-versus-foreign allocation decision, whether repatriation is a real prospect, and what it would take for capital to come home in size.

For professional readers tracking fixed income and FX market dynamics, this episode offers a rare look inside the institutional decision-making that drives these important capital flows. Listeners are encouraged to tune in to hear the guest's specific analysis and judgments on these issues.

Fontes:omny.fm

Pesquisa

WeMakeDevs Founder Kunal Kushwaha: As AI Takes on More Implementation Work, Engineers Are Advised to Run Large Models Locally to Deepen Technical Understanding

Material
2026-10-21 08:00 GMT+8

If you ask a developer about their workflow, they can tell you how much faster the latest model is, what their coding agent can now handle, and which new tool has improved the way they work. Kunal Kushwaha has a follow-up question.

"What about you?"

Kushwaha’s answer starts with the work that still belongs to the engineer. He believes that regardless of the tools used, engineers need to understand the problem well enough to choose an approach and explain the reasoning behind their decisions; they remain responsible for what ships. As AI takes on more implementation work, where does this judgment come from? He also pointed to a practical direction: learning to run large language models locally.

The purpose of this exercise is not to replace API calls, but to expose the infrastructure behind the interface. Kushwaha said that running a model locally and then exploring fine-tuning and what it would take to run at scale allows engineers to access how the model runs, how its components communicate, and the deployment choices involved. He compared it to learning Kubernetes: an engineer may use managed services in production, but still benefits from having built and operated a small cluster themselves—that experience helps them understand what the managed service is doing on their behalf. The same principle applies when a team chooses how to deploy models: APIs, company-owned hardware, and managed platforms each have different tradeoffs, and hands-on local setup provides more context to ask better questions.

Why do fundamentals still matter? Kushwaha’s answer is direct: you learn how to escape the hype cycle. Understanding the basics makes it easier to ask what a tool actually changes, which part of the system it helps, and where its limitations lie. Without that grounding, it is easy to mistake a better demo for a better solution to the problem in front of you. He does not believe engineers need to know everything—he uses Kubernetes deeply himself and still encounters unfamiliar details in Kubernetes, Linux, and networking. Hands-on experience provides enough context to investigate the unknown and make informed decisions.

On the degree of automation in code review, Kushwaha diverged from the interviewer. He expects agents to handle more review work, with humans eventually stopping code review altogether; however, he believes the level of risk should determine attention to changes—if an agent produces a fix during a major promotion period, he still wants someone to check it. The bottom line both agree on is: responsibility does not shift to the agent. If something goes wrong, you are the one who is responsible.

It should be noted that these suggestions come from a practitioner's personal experience and analogical reasoning, not a systematically verified optimal path. Running large language models locally requires corresponding hardware and time investment; as AI tools iterate rapidly, whether individual skill accumulation can keep pace with the speed of structural change is a question this conversation did not answer.

Fontes:CodeRabbit

Pesquisa

Colossus Podcast Analyzes Affirm: How AI Transaction-Level Underwriting and Merchant Networks Build Moats

Material
2026-10-22 16:00 GMT+8

Colossus's Business Breakdowns podcast released its latest episode titled "Affirm: The Internet’s American Express" on October 22, 2026. The content specifically breaks down Affirm, the buy now, pay later (BNPL) company founded by PayPal co-founder Max Levchin. According to the episode description, Affirm currently processes more than $50 billion in annual purchase volume and continues to grow over 30%, even after its stock fell more than 90% following its IPO.

The guest joining the host is an investor named Fraser, whose background is described as having spent time with Max Levchin and studied the business closely. The episode outline indicates that core discussion topics include how Affirm makes money, why its transaction-level underwriting differs from traditional credit cards, and how its merchant relationships create powerful network effects. Additionally, the show explores the company's resilience through rising interest rates, its competition with Klarna, and why Affirm may be better understood as the American Express of the internet rather than a traditional lender.

For readers focused on AI industry investment and the fintech sector, this podcast provides a structured framework for industry analysis. Although a full transcript has not been obtained, based on the official timestamps and topic list, listeners can focus on specific segments such as "AI and the advantage of transaction-level underwriting," "The power of negative CAC," and "How AI agents could transform payments." These topics directly relate to the current application of AI technology in vertical financial domains and its potential impact on traditional credit models.

Fontes:colossus.com

Pesquisa

Latent Space AINews: OpenAI and Wiz Use AI Models to Discover PHP Core Vulnerabilities, Triggering Global Coordinated Remediation

Tópico · PHP2Shell漏洞Material
2026-10-22 15:42 GMT+8

This week's Latent Space AINews roundup (covering October 20–21, 2026) places the headline focus on the recent PHP2Shell vulnerability incident. The report notes that OpenAI's "Patch the Planet" project and security firm Wiz acted as two independent parties, leveraging advanced AI models such as GPT-5.6 Sol and Gemini 4 Argon to discover two critical vulnerabilities (CVE-2026-109618 and CVE-2026-109619) in PHP's internal handling of out-of-memory conditions and fatal errors. These flaws allowed unauthenticated remote code execution against major CMS platforms like WordPress, Drupal, and Joomla, with some researchers describing them as the most consequential vulnerabilities they have helped discover.

The report emphasizes that this was a typical multi-party collaboration rather than a single-vendor victory. OpenAI described an approximately two-month coordinated remediation period involving close cooperation among PHP maintainers, researchers, security providers, and even governments. Cloudflare confirmed via its CTO that initial WAF mitigations were deployed before the scheduled patch release, covering known exploit vectors and available to all customers, but explicitly stated these rules were a stopgap and that patching remained mandatory. Meanwhile, the U.S. National Security Agency (NSA) issued warnings advising operators to prepare for immediate updates and consider taking systems offline if timely mitigation was not possible.

Beyond the core security incident recap, this issue aggregates several important AI product and infrastructure developments. On the model front, Recursive released Flint, an Apache-2.0 licensed local coding MoE model, claiming it was assisted by their automated research system Eureka Machine and achieved significant results on benchmarks like SWE-bench. Anthropic launched a research preview of Claude Duet, a proactive personal assistant powered by Opus 5.5, supporting asynchronous task execution and financial operation approval boundaries.

On the hardware and supply chain level, citing Reuters, the report notes that ByteDance, Alibaba, and Tencent have placed over $35 billion in Nvidia chip orders, primarily for professional cards like the RTX PRO 5500, though actual delivery volumes may be significantly lower due to production constraints. Additionally, Samsung is reportedly sold out of its entire 2027 HBM capacity, expected to drive up HBM prices and consequently increase manufacturing costs for GPUs and AI accelerators. These dynamics collectively illustrate the current AI industry's landscape regarding upgraded defensive capabilities, localized model deployment, and tight compute supply chains.

Fontes:latent.space

Pesquisa

Anthropic Releases Research Preview of Claude Duet, a Persistent Autonomous AI Agent

Material
2026-10-22 15:14 GMT+8

US artificial intelligence company Anthropic announced the research preview of Claude Duet on October 21. This persistent autonomous AI agent is built on the Claude Opus 5.5 model and is designed to handle complex tasks even when users step away from their devices. The feature is currently available only to US subscribers on Pro and Max plans via a waitlist, with Team and Enterprise plans excluded for now.

The core capability of Claude Duet lies in its ability to operate independently of local hardware. It runs within a cloud-based virtual environment, continuing to execute tasks such as scheduling meetings, building dashboards, and generating financial forecasts even if the user's PC or smartphone is powered off. The system checks linked accounts periodically and notifies users upon task completion or when a decision is required. It supports over 6,000 connectors and plugins, assisting with drafting Gmail replies or reviewing unpaid invoices in QuickBooks.

To mitigate risks, Anthropic has implemented strict operational boundaries. The agent requires explicit user approval before sending messages, making payments, signing documents, or deleting data. High-risk actions such as securities trading or transferring funds between accounts are prohibited entirely, even with user consent. As this remains a research preview, the company acknowledges potential errors in interpreting conversation context or assessing urgency, noting that background verification processes also count against usage limits.

Fontes:PC Watch

Pesquisa

Japan's LSTC and Partners Secure NEDO Grant for 2nm Medical AI Chip Project Targeting 5x GPU Speed

Tópico · LSTC医用AI芯片项目Visão rápida
2026-10-22 23:30 GMT+8

On October 20, 2026, the Research Association of Advanced Semiconductor Technology Center (LSTC), Hokkaido University, Fujifilm, GE Healthcare Japan, and imec announced that their joint project, "Research and Development of High-Speed, Low-Power Medical Image Processing Systems Using Inference-Specific 2nm Semiconductors," was selected for funding by Japan's New Energy and Industrial Technology Development Organization (NEDO). The initiative aims to address power consumption and latency barriers in medical AI while creating use cases for advanced semiconductors manufactured in Japan.

The team plans a two-stage approach: first developing and optimizing three core AI systems—CT image reconstruction, organ segmentation, and real-time surgical support—on IBM's "Spyre Accelerator," then migrating them to 2nm inference-specific AI chips. The project targets achieving at least 5x faster image processing and 10x higher energy efficiency compared to current GPUs, with final validation planned at Hokkaido University Hospital.

Fontes:EE Times Japan

Pesquisa

Canadian Real-Time OS Company QNX Proposes OS-Level Control for Physical AI Uncertainty and Launches Free Developer Program

Tópico · QNX物理AI安全Visão rápida
2026-10-22 13:12 GMT+8

Canadian real-time operating system company QNX held a technology forum in Seoul on October 22, 2026, proposing a solution to control AI uncertainty at the operating system (OS) level in response to growing safety concerns in robotics and autonomous driving. Kim Sung-chul, General Manager of QNX Korea, stated that as humanoid robots and autonomous mobile robots (AMRs) become more widespread, risks of personal injury from unexpected behaviors are increasing, requiring more rigorous pre-action safety judgments for enterprises.

Justin Moon, Vice President of Core Product Engineering at QNX, pointed out that unlike digital AI which only processes data in software, physical AI acts directly in the real world, necessitating trust built from the ground up. He emphasized three conditions: safety, reliability, and determinism, defining determinism as the property of producing the same output for the same input within a fixed time. Moon argued that unpredictable timing is not just a performance issue but a safety one. QNX positions itself as the execution layer between AI decision-making and physical action, ensuring the entire system responds appropriately regardless of how the AI model operates.

Addressing industry claims that "AI runs only on Linux," Moon claimed that benchmark tests conducted by QNX under identical silicon, models, and workloads showed distinct differences, inviting external verification. Additionally, QNX showcased "Alloy Core," a platform co-developed with German automotive software company Vector, combining QNX's safety-certified OS with Vector's middleware. Recently, this platform secured QNX's largest-ever supply contract with a joint venture between Daimler Truck and Volvo Group.

To expand its developer ecosystem, QNX launched the "QNX Everywhere" program, offering QNX OS, hypervisors, and educational content for free to non-commercial users. Ryu Min-hee, Senior Manager at QNX Korea, stated this initiative aims to strengthen the ecosystem compared to Linux. According to Andy Gric, Director of Developer Ecosystems at QNX, the program has provided 30,000 development seats, partnered with over 100 universities worldwide, and ported 2,500 open-source packages to QNX. QNX plans to start its Asian academic outreach in South Korea before expanding to Japan and China, currently collaborating with Jeong Gu-min, a professor at Konkuk University, to establish a "QNX Center" introducing the OS into undergraduate and graduate education.

Fontes:zdnet.co.kr

Pesquisa

Microsoft says China-linked hackers used AI penetration testing tool ARTEX to compromise ConnectWise and target hundreds of organizations

Estrutural
2026-10-22 02:23 GMT+8

In a research report published on October 21, 2026, Microsoft Security Blog attributed a supply chain attack against IT services firm ConnectWise and a subsequent campaign targeting hundreds of global organizations to Storm-0940, a China-state aligned threat actor. The report states that since September 2026, the group has been using ARTEX, an open-source agentic AI penetration testing framework, to automate its offensive workflows.

ARTEX is a multi-agent LLM system capable of autonomously executing penetration tests from reconnaissance through data exfiltration with minimal human interaction. Microsoft observed that Storm-0940 leveraged ARTEX to accelerate scanning, password spraying, and OAuth application abuse. In multiple instances, the time from initial targeting to obtaining privileged access to an organization’s cloud infrastructure was less than three hours. This machine-speed activity generated significant log volume but left defenders with very little time to coordinate a response.

The campaign targeted sectors including aerospace, defense, automotive, energy, financial services, government, legal, manufacturing, media, semiconductor, and software. While most attempts were unsuccessful, Storm-0940 achieved privileged cloud persistence at many organizations in a brief period. Microsoft emphasized that these activities did not result from vulnerabilities in Microsoft’s own products or services, but rather from the abuse of previously granted third-party application access.

ConnectWise had previously issued an advisory on October 16 regarding unauthorized access attempts associated with its Cloud Backup environment. Microsoft advised customers to review service principal sign-ins, investigate unexpected application credential or privilege changes, and follow ConnectWise’s containment guidance.

Fontes:Microsoft Security Blog

Pesquisa

Harvard and European Commission Measurement Finds: Fragmentation of European Regulation Reduces Competitiveness in Complex Research Fields, with Exceptions Only for Green Tech and Semiconductors

Tópico · 欧盟AI监管碎片化Material
2026-10-22 14:00 GMT+8

In an article published on October 22, 2026, investment blogger Joachim Klement cited research by Balland et al. (2026) from a team at Harvard and the European Commission, offering a structural explanation for Europe's technological competitiveness: the fragmentation of regulatory and funding systems increases cross-institutional collaboration costs, and these costs amplify as the complexity of the research field increases.

European countries have different regulatory rules and funding agencies. Klement points out that a researcher working in Germany attempting to collaborate with peers in Spain or the UK must navigate a jungle of institutions, rules, and regulations. In contrast, academic researchers and technology development enterprises in the United States face essentially one unified system of grant-making institutions and one unified regulatory framework.

The key is that this difference does not constitute an obstacle in research that can be conducted within a single institution. However, when research enters fields such as AI, quantum computing, and biotechnology, requiring increasingly more collaborators with diverse expertise, the fragmented institutional environment translates into actual collaboration friction. Klement concludes that the less fragmented research environment in the United States provides a structural advantage over Europe in complex fields.

The research team led by Balland quantified this degree of fragmentation by measuring collaboration patterns in patent filings and scientific publications, and scored the complexity of various research fields. The measurements revealed a clear pattern: the collaborative advantage of the United States becomes more pronounced in fields with higher complexity.

Specifically, Europe holds a relative advantage in engine and pump technology, telecommunications technology, green technology, and semiconductors; the United States leads in medical technology, biotechnology, broader life sciences, and computer technology. Among all assessed complex research fields, green technology and semiconductors are the only two directions where Europe maintains a lead.

The study authors conclude that the high degree of fragmentation in the European research environment is becoming an increasingly significant obstacle in competition for cutting-edge research in the 21st century. For investors, this finding suggests that the institutional environment is a structural factor that needs to be considered when assessing the long-term competitiveness of European technology companies.

It should be noted that the study presents the correlation between fragmentation and competitiveness by measuring collaboration patterns and scoring field complexity; it does not claim that fragmentation is the sole explanatory factor.

Fontes:klementoninvesting.substack.com

Pesquisa
Próxima →
Entrar / Cadastrar
LeituraPesquisaRadarFramework de investimento